Customer Cost Forecasting Through Patterns Learning of Optimal Capacity Allocation

Production, logistics, and service systems involve multiple customers being served by shared resources. Once either a process or a service concerning several customers is optimized, business sustainability and profit opportunities become a matter of contracted commercial offers and business inherent risks sharing. Commercial offers are indeed intended to cover the expected costs, be competitive without reducing net margin, and robust to risks resulting from the uncertainties affecting the business. In this study we develop a framework that integrates an operational research model providing optimal capacity allocation with a learning model performing customer cost estimation. This estimator relies over a set of input features related to each customer. The aim is to provide a cost forecasting framework for contract management that considers a fair allocation of costs to customers. Validation of the approach is done with real case instances of a pick-up and delivery routing model dealing with reverse logistic operations. The authors propose a new formulation of the considered routing problem and show the effectiveness of the proposed method